A high-speed servo drive FOC-MPC fusion control method and system based on GaN devices

Abstract

This invention provides a high-speed servo drive FOC-MPC fusion control method and system based on GaN devices, relating to the field of industrial robot control technology. The method includes: selecting a final voltage space vector from the basic voltage space vector of a GaN device three-phase full-bridge inverter based on a reference voltage vector in a stationary coordinate system; determining the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter within the current switching cycle based on the final voltage space vector; driving each GaN switch in the GaN device three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range via a gate drive circuit according to the final switching state combination; and outputting a drive voltage to the stator winding of a permanent magnet synchronous motor through the midpoint of the bridge arm according to the switching state switching sequence of each GaN switch under the final switching state combination. This invention eliminates the delay of the modulation stage, achieving delay-free voltage command application.

Description

Technical Field

[0001] This invention relates to the field of industrial robot control technology, and in particular to a high-speed servo drive FOC-MPC fusion control method and system based on GaN devices. Background Technology

[0002] Servo drive systems are the core execution units for achieving precision motion control in industrial automation equipment. Existing servo drives generally employ a field-oriented control (FOC) strategy, achieving closed-loop control through a cascaded three-loop structure of position, speed, and current loops. Specifically, the current loop generates a reference voltage vector based on the deviation between the current command and the feedback current via proportional-integral regulation. This reference voltage vector is then converted into a PWM duty cycle signal for the inverter's power switch transistors via space vector pulse width modulation (SVPWM) or sinusoidal pulse width modulation (SPWM) stages, ultimately controlling the power switch transistors to drive the motor.

[0003] However, in the aforementioned traditional FOC control strategy, the modulation stage introduces a calculation delay and modulation delay of one to two beats between the output of the current loop regulator and the actual output voltage of the inverter. This prevents the actual voltage command of the current loop from acting on the motor stator windings without delay within the current switching cycle. This delay directly limits the effective bandwidth of the current loop, causing the actual current to fail to quickly track changes in the current command when the servo system performs high-frequency start / stop or rapid position step changes. This results in a lag in torque response, manifested as increased position following error and decreased dynamic positioning accuracy. Summary of the Invention

[0004] This invention provides a high-speed servo drive FOC-MPC fusion control method and system based on GaN devices, which eliminates the delay of the modulation stage and realizes the application of voltage commands without delay.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, a high-speed servo drive FOC-MPC fusion control method based on GaN devices is provided, the method comprising: Step 1: Obtain the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, obtain the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system. Step 2: Obtain the speed command based on the actual rotor speed and position command; obtain the q-axis current command based on the speed deviation between the speed command and the actual rotor speed. Step 3: Based on the q-axis current command, the actual d-axis current, and the actual q-axis current, perform current loop adjustment on the d-axis current deviation and the q-axis current deviation to obtain the d-axis reference voltage vector and the q-axis reference voltage vector; based on the d-axis reference voltage vector and the q-axis reference voltage vector, obtain the reference voltage vector in the stationary coordinate system through inverse coordinate transformation; Step 4: Select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. Step 5: Based on the final switching state combination, drive each GaN switch in the GaN device three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit. Step 6: Based on the switching sequence of each GaN switch under the final switching state combination, output drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm.

[0006] Secondly, a high-speed servo-driven FOC-MPC fusion control system based on GaN devices includes: The acquisition module is used to acquire the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, it obtains the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system; based on the actual rotor speed and position command, it obtains the speed command; based on the speed deviation between the speed command and the actual rotor speed, it obtains the q-axis current command. The first calculation module is used to perform current loop adjustment on the d-axis current deviation and q-axis current deviation according to the q-axis current command, the actual d-axis current and the actual q-axis current, to obtain the d-axis reference voltage vector and the q-axis reference voltage vector; and to obtain the reference voltage vector in the stationary coordinate system by inverse coordinate transformation based on the d-axis reference voltage vector and the q-axis reference voltage vector. The second calculation module is used to select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; and to determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. The processing module is used to drive each GaN switch in the three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit according to the final switching state combination; and to output a drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm according to the switching state switching sequence of each GaN switch under the final switching state combination.

[0007] Thirdly, a computing device includes: Multiple processors; A storage device for storing a plurality of programs, which, when executed by a plurality of processors, cause the plurality of processors to implement the method.

[0008] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0009] The above-described solution of the present invention has at least the following beneficial effects: The FOC-MPC fusion architecture eliminates the computational delay introduced by the traditional modulation stage, enabling the voltage command of the current loop to act on the motor stator winding without delay within one switching cycle. As a result, the bandwidth of the current loop is significantly improved, and the current tracking accuracy is significantly improved compared to the traditional FOC plus modulation scheme.

[0010] The high switching frequency significantly shortens the duration of each switching cycle, thereby increasing the frequency at which the MPC module completes voltage vector optimization and switching state updates within each cycle, and greatly reducing the discretization error of current tracking. Simultaneously, at high switching frequencies, the higher harmonic frequencies in the inverter output voltage are much higher than the motor fundamental frequency, effectively suppressing the switching ripple in the motor stator current, resulting in an output current waveform closer to a sine wave and a significant reduction in torque ripple. Traditional silicon-based IGBT solutions are limited by their lower switching frequencies, and even with the introduction of MPC algorithms, they cannot achieve the same current tracking performance at the same frequency.

[0011] This application establishes a resonant network between the DC bus and the midpoint of the bridge arm in a three-phase full-bridge inverter using GaN devices. Soft-switching commutation is achieved by performing turn-on operations at the zero-voltage moment when the drain-source voltage of each GaN switch naturally resonates to zero, and by performing turn-off operations at the zero-current moment when the drain-source current naturally resonates to zero. Soft-switching commutation effectively reduces the turn-on and turn-off losses of GaN devices under high-frequency switching conditions, enabling the inverter system efficiency to reach a high level under rated load and solving the problem of increased losses due to high-frequency operation. Compared to the limited system efficiency achievable only at lower switching frequencies by traditional silicon-based IGBT solutions, this invention achieves an increase in efficiency while increasing the frequency several times over.

[0012] GaN enhancement-mode high electron mobility transistors (GMT-HMTs) have low gate threshold voltages, making them susceptible to gate voltage erroneous turn-on due to rapid drain-source voltage changes during high-speed switching, which can lead to bridge arm shoot-through faults. This invention addresses this issue by combining an active Miller clamping circuit with a negative voltage direct-drive gate drive circuit. During gate turn-off, the gate voltage is actively clamped to a preset negative voltage level, effectively suppressing gate voltage erroneous turn-on caused by the Miller effect. Simultaneously, a positive voltage drive pulse is provided during gate turn-on to ensure reliable conduction, and a negative voltage turn-off pulse is provided during turn-off to accelerate turn-off and enhance anti-interference capabilities, ensuring the safe and reliable operation of GaN devices under high switching frequencies and high voltage change rates. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating a high-speed servo drive FOC-MPC fusion control method based on GaN devices, provided by an embodiment of the present invention.

[0014] Figure 2 This is a schematic diagram of a high-speed servo drive FOC-MPC fusion control system based on GaN devices provided in an embodiment of the present invention. Detailed Implementation

[0015] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0016] like Figure 1 As shown, an embodiment of the present invention proposes a high-speed servo drive FOC-MPC fusion control method based on GaN devices, the method comprising the following steps: Step 1: Obtain the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, obtain the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system. Step 2: Obtain the speed command based on the actual rotor speed and position command; obtain the q-axis current command based on the speed deviation between the speed command and the actual rotor speed. Step 3: Based on the q-axis current command, the actual d-axis current, and the actual q-axis current, perform current loop adjustment on the d-axis current deviation and the q-axis current deviation to obtain the d-axis reference voltage vector and the q-axis reference voltage vector; based on the d-axis reference voltage vector and the q-axis reference voltage vector, obtain the reference voltage vector in the stationary coordinate system through inverse coordinate transformation; Step 4: Select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. Step 5: Based on the final switching state combination, drive each GaN switch in the GaN device three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit. Step 6: Based on the switching sequence of each GaN switch under the final switching state combination, output drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm.

[0017] In this embodiment, in the traditional FOC control strategy, the reference voltage vector output by the current loop must be converted into a PWM duty cycle signal through a space vector pulse width modulation or sinusoidal pulse width modulation stage. This modulation stage introduces a calculation delay and modulation delay of one to two beats between the output of the current loop regulator and the actual output voltage of the inverter, which causes the actual voltage command of the current loop to be unable to act on the motor stator winding without delay in the current switching cycle, thus limiting the effective bandwidth of the current loop.

[0018] This invention, after generating a reference voltage vector in the stationary coordinate system in the FOC current loop, bypasses the traditional modulation stage. Instead, it directly inputs this reference voltage vector into the MPC module. The MPC module then traverses all the basic voltage space vectors of the GaN device three-phase full-bridge inverter within the current switching cycle. With the goal of minimizing the tracking error between the predicted current and the reference current, it selects the optimal voltage space vector online and directly determines the final switching state combination of each GaN switch. This FOC-MPC fusion architecture eliminates the computational delay introduced by the traditional modulation stage, achieving zero-delay application of the current loop voltage command within the current switching cycle, thereby significantly improving the effective bandwidth and current tracking accuracy of the current loop.

[0019] This invention, while maintaining the three-loop cascaded structure of position loop, speed loop, and current loop under the FOC framework, embeds the MPC module at the output of the current loop to replace the traditional modulation stage. This allows the output of the FOC's pre-stage regulator to be directly connected to the online optimization of the MPC. The outputs of the position loop and speed loop serve as the given commands for the current loop, and the proportional-integral regulation output of the current loop serves as the reference voltage vector for the MPC. The MPC directly outputs the optimal switching state combination of the switching transistors to drive the inverter.

[0020] This invention combines a GaN-based three-phase full-bridge inverter with an MPC module: the high switching frequency significantly shortens the duration of each switching cycle, thereby increasing the frequency at which the MPC module completes voltage vector optimization and switching state updates within each cycle, and greatly reducing the discretization error of current tracking; at the same time, the high-order harmonic frequencies contained in the inverter output voltage at the high switching frequency are much higher than the fundamental frequency of the motor, effectively suppressing the switching ripple in the motor stator current, making the output current waveform closer to a sine wave, and significantly reducing torque ripple.

[0021] GaN enhancement-mode high electron mobility transistors (GMT-HMTs) have low gate threshold voltages, making them susceptible to gate voltage erroneous turn-on due to rapid drain-source voltage changes during high-speed switching, which can lead to bridge arm shoot-through faults. This invention addresses this issue by combining an active Miller clamping circuit with a negative voltage direct-drive gate drive circuit. During gate turn-off, the gate voltage is actively clamped to a preset negative voltage level, effectively suppressing gate voltage erroneous turn-on caused by the Miller effect. Simultaneously, a positive voltage drive pulse is provided during gate turn-on to ensure reliable conduction, and a negative voltage turn-off pulse is provided during turn-off to accelerate turn-off and enhance anti-interference capabilities, ensuring the safe and reliable operation of GaN devices under high switching frequencies and high voltage change rates.

[0022] This invention establishes a resonant network between the DC bus and the midpoint of the bridge arm in a three-phase full-bridge inverter using GaN devices. It achieves soft-switching commutation by performing turn-on operations at the zero-voltage moment when the drain-source voltage of each GaN switch naturally resonates to zero, and by performing turn-off operations at the zero-current moment when the drain-source current naturally resonates to zero. The introduction of soft-switching technology effectively reduces the turn-on and turn-off losses of GaN devices under high-frequency switching conditions, enabling the inverter system to achieve higher operating efficiency under rated load and solving the problem of increased losses caused by high-frequency operation.

[0023] In a preferred embodiment of the present invention, step 1 involves obtaining the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, obtaining the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system, including: Step 11: Based on the current sensors installed in the three-phase stator winding circuit of the permanent magnet synchronous motor, the instantaneous values ​​of the three-phase stator current are acquired at a sampling rate synchronized with the switching frequency of the GaN device three-phase full-bridge inverter. The rotor mechanical angular position is then acquired based on the encoder installed at the rotor shaft end of the permanent magnet synchronous motor. Specifically, this includes: In this step, the execution process for current sampling and position sampling is as follows: Based on three current sensors connected to the stator winding circuits of phases A, B, and C of the permanent magnet synchronous motor, the instantaneous current values ​​flowing through each phase winding are synchronously collected at the sampling trigger time, which is aligned with the start time of the current switching cycle of the GaN device three-phase full-bridge inverter. The instantaneous values ​​of the stator current of phase A, phase B, and phase C are obtained, and the three together constitute the instantaneous value of the three-phase stator current.

[0024] Simultaneously, based on the position encoder installed at the rotor shaft end of the permanent magnet synchronous motor, the angle pulse count value output by the encoder is read at the sampling trigger moment. According to the proportional relationship between the angle pulse count value and the encoder line count, the current rotor position expressed in mechanical angles is calculated, thus obtaining the rotor mechanical angular position.

[0025] Step 12: Based on the instantaneous values ​​of the three-phase stator currents, obtain the α-axis current components and β-axis current components in the α-β stationary coordinate system through Clarke transformation; specifically including: In this step, the execution process of the Clarke transform is as follows: Based on the instantaneous stator current values ​​of phase A, phase B, and phase C output in step 11, these instantaneous stator current values ​​are mapped from the three-phase stationary coordinate system to a two-phase orthogonal stationary coordinate system. During the mapping process, the axis of phase A winding is used as the reference direction of the α-axis, and the direction that leads the α-axis by 90 electrical degrees in space is used as the direction of the β-axis. Based on the spatial distribution of the three-phase currents differing by 120 electrical degrees, the projection of the instantaneous A-phase current value onto the α-axis is linearly combined with the projections of the instantaneous B-phase and C-phase current values ​​onto the α-axis to obtain the α-axis current component. Simultaneously, the projections of the instantaneous B-phase and C-phase current values ​​onto the β-axis are linearly combined to obtain the β-axis current component. This transformation maintains the principle of amplitude invariance, meaning the amplitude of the synthesized current vector is equal before and after the transformation.

[0026] Step 13: Based on the α-axis current component and the β-axis current component, and the electrical angle calculated based on the rotor mechanical angular position, the actual d-axis current and actual q-axis current in the dq rotating coordinate system are obtained through Parker transformation; specifically including: In this step, the execution process of the Parker transform is as follows: Based on the rotor mechanical angular position output in step 11, the electrical angle is calculated. The calculation method is as follows: multiply the rotor mechanical angular position by the number of pole pairs of the permanent magnet synchronous motor, and perform a reduction process with a period of 360 degrees of electrical angle to ensure that the electrical angle is always within the range of 0 degrees to 360 degrees of electrical angle, thus obtaining the electrical angle corresponding to the current rotor position.

[0027] Based on the α-axis and β-axis current components output in step 12, and combined with the electrical angle, the current vector in the α-β stationary coordinate system is transformed to the dq rotating coordinate system, which rotates synchronously with the rotor magnetic field. During the transformation, the direction of the rotor permanent magnet flux linkage is taken as the positive direction of the d-axis, and the direction that leads the d-axis by 90 degrees in space is taken as the positive direction of the q-axis. The product of the α-axis current component and the cosine of the electrical angle is added to the product of the β-axis current component and the sine of the electrical angle to obtain the actual d-axis current. The product of the β-axis current component and the cosine of the electrical angle is subtracted from the product of the α-axis current component and the sine of the electrical angle to obtain the actual q-axis current.

[0028] This step yields the actual d-axis current and the actual q-axis current, which respectively characterize the excitation component of the stator current vector in the direction of the rotor magnetic field and the torque component orthogonal to it.

[0029] Step 14: Obtain the actual rotor speed by performing a differential calculation of the rotor's mechanical angular position with respect to time.

[0030] In this step, the execution process of speed calculation is as follows: Based on the rotor mechanical angle position at the current sampling moment output in step 11, and the previous rotor mechanical angle position acquired in the previous switching cycle in step 11, the mechanical angle difference between the two position sampling values ​​is calculated. Simultaneously, based on the preset switching frequency of the GaN device three-phase full-bridge inverter, the time interval between two adjacent samplings is determined; this time interval is equal to the duration of the switching cycle.

[0031] By dividing the mechanical angle difference by the time interval, a first-order backward difference operation is performed on the rotor's mechanical angular position to obtain the rotor's current mechanical angular velocity. Based on this rotor mechanical angular velocity and the conversion relationship between radians and rotational speed units, the mechanical angular velocity is converted into a rotational speed value in revolutions per minute (rpm), yielding the actual rotor speed. The formula for calculating the mechanical angular velocity using the first-order backward difference is as follows: ; in, Indicates the first The rotor mechanical angular velocity is calculated at each sampling time (i.e., the current switching cycle), and the unit is radians per second (rad / s). Indicates the first The rotor mechanical angular position is obtained by the encoder at each sampling time (current switching cycle), and the unit is radians (rad). Indicates the first The rotor mechanical angular position, measured in radians (rad), is obtained by the encoder at each sampling time (i.e., the previous switching cycle). This represents the sampling time interval, i.e., the switching cycle of a GaN device three-phase full-bridge inverter, measured in seconds (s). Its value is equal to the reciprocal of the switching frequency. ).

[0032] To obtain the rotor mechanical angular velocity in radians per second Then, convert it to a unit of rotational speed, i.e., revolutions per minute (r / min, i.e., rpm). The conversion formula is: ;Right now, ; in, Indicates the first The actual rotor speed at each sampling time, in revolutions per minute (r / min). This represents the rotor mechanical angular velocity calculated in the first stage, in radians per second (rad / s). This represents a time conversion factor where each minute is 60 seconds, meaning 1 minute equals 60 seconds. This indicates the number of radians corresponding to one revolution of the rotor, i.e., 1 revolution = radian.

[0033] By combining the above first-order backward differential calculation of mechanical angular velocity and rotor mechanical angular velocity, we obtain a complete formula for directly calculating the actual rotor speed from the rotor's mechanical angular position: ; in, and These are the rotor mechanical angular positions (rad) for the current and previous switching cycles, respectively. The switching period (s) is the time interval between the two switches. The current rotational speed (r / min) is calculated.

[0034] In a preferred embodiment of the present invention, step 2 includes: Step 21: Based on the actual rotor speed and the position command received from the upper-level controller, perform position loop proportional adjustment on the position deviation between the position command and the actual rotor position to obtain the speed command.

[0035] Specifically, in this step, the execution process of the position loop proportional adjustment is as follows: Based on the rotor mechanical angular position acquired and converted by the encoder in step 11, the actual rotor position is determined. Simultaneously, a position command is received from the upper-level controller, which is the desired rotor mechanical angular position generated by the upper-level controller based on the target motion trajectory planned by the servo system.

[0036] Based on this, the position command is subtracted from the actual rotor position, and the difference between the two is calculated to obtain the position deviation.

[0037] Based on this, a position loop proportional adjustment is performed according to the position deviation. The specific adjustment algorithm is as follows: the position deviation is multiplied by a preset position loop proportional gain coefficient, and the resulting product is the amplitude of the speed command. The value of the position loop proportional gain coefficient is pre-tuned based on the rated speed, load inertia, and position response bandwidth requirements of the permanent magnet synchronous motor. When the position deviation is positive, the resulting speed command is a positive direction speed command; when the position deviation is negative, the resulting speed command is a negative direction speed command.

[0038] During the position loop proportional adjustment process, the amplitude of the speed command is limited, that is: the absolute value of the speed command is compared with a preset speed command upper limit threshold; if the absolute value of the speed command exceeds the speed command upper limit threshold, the speed command is clamped to a limit value with the same sign as the speed command upper limit threshold; if it does not exceed the limit, the original value is kept unchanged. After this step, the speed command is obtained.

[0039] Step 22: Based on the speed deviation between the speed command and the actual rotor speed, perform speed loop proportional-integral adjustment to obtain the q-axis current command; at the same time, set the d-axis current command to zero.

[0040] Specifically, in this step, the execution process of the speed loop proportional-integral control is as follows: Based on the speed command output in step 21 and the actual rotor speed output in step 14, the speed command is subtracted from the actual rotor speed, and the difference between the two is calculated to obtain the speed deviation.

[0041] Based on this, speed loop proportional-integral adjustment is performed according to the speed deviation. The speed loop proportional-integral adjustment includes two parts: a proportional calculation channel and an integral calculation channel.

[0042] In the proportional calculation channel, the speed deviation is multiplied by a preset speed loop proportional gain coefficient to obtain the proportional adjustment component. The value of the speed loop proportional gain coefficient is determined in advance based on the torque constant, moment of inertia, and desired bandwidth of the permanent magnet synchronous motor.

[0043] In the integral operation channel, the speed deviation is accumulated and integrated. Specifically, within each switching cycle, the speed deviation of the current switching cycle is multiplied by a preset speed loop integral gain coefficient, and then multiplied by the time step of the switching cycle to obtain the integral increment of the current switching cycle; the integral increment is added to the integral accumulation value of the previous switching cycle to obtain the integral adjustment component of the current switching cycle. The value of the speed loop integral gain coefficient is pre-tuned based on the steady-state accuracy requirements of the speed loop and the integral saturation constraint.

[0044] Accordingly, the proportional adjustment component is added to the integral adjustment component to obtain the q-axis current command.

[0045] After receiving the q-axis current command, the amplitude of the q-axis current command is limited. Specifically, the absolute value of the q-axis current command is compared with a preset upper threshold value for the q-axis current command. If the absolute value of the q-axis current command exceeds the upper threshold value, the q-axis current command is clamped to a limit value with the same sign as the upper threshold value. If it does not exceed the threshold value, the original value remains unchanged. The upper threshold value for the q-axis current command is preset based on the rated current and demagnetizing current constraints of the permanent magnet synchronous motor.

[0046] While the speed loop proportional-integral regulation is being executed, the d-axis current command is directly set to zero. The basis for setting the d-axis current command to zero is that, for surface-mounted permanent magnet synchronous motors, a zero d-axis current control strategy (i.e., id=0 control) is adopted, so that the stator current vector is used entirely to generate electromagnetic torque, thereby achieving maximum torque-to-current ratio operation. After this step, the q-axis current command and the d-axis current command are obtained.

[0047] This step introduces both proportional and integral control into the speed loop, enabling it to respond quickly to changes in speed deviation (guaranteed by the proportional channel) and eliminate steady-state speed error (guaranteed by the integral channel). This ensures that the servo system tracks the speed command without steady-state error during steady-state operation. Simultaneously, setting the d-axis current command to zero simplifies the current distribution strategy, making the q-axis current command linearly related to the electromagnetic torque, facilitating tracking control by the current loop.

[0048] In a preferred embodiment of the present invention, step 3 includes: Step 31: Calculate the q-axis current deviation based on the difference between the q-axis current command and the actual q-axis current, and calculate the d-axis current deviation based on the difference between the d-axis current command and the actual d-axis current, wherein the d-axis current command is zero.

[0049] In this step, the calculation process for the d-axis current deviation and the q-axis current deviation is as follows: Based on the d-axis current command output in step 22 and the actual d-axis current output in step 13, the d-axis current command is subtracted from the actual d-axis current, and the difference between the two is calculated to obtain the d-axis current deviation. Since the d-axis current command is set to zero, the d-axis current deviation is numerically equal to the negative of the actual d-axis current, that is: ; in, For the first d-axis current deviation per switching cycle For the first The d-axis current command for each switching cycle (constantly zero). For the first The actual d-axis current for each switching cycle. All parameters are in amperes (A). When the actual d-axis current is positive, the d-axis current deviation is negative, indicating that the actual excitation current is too large and needs to be suppressed by a negative voltage. When the actual d-axis current is negative, the d-axis current deviation is positive, indicating that the actual excitation current is too small (in the demagnetizing direction) and needs to be compensated by a positive voltage.

[0050] Based on this, according to the q-axis current command output in step 22 and the actual q-axis current output in step 13, the q-axis current command is subtracted from the actual q-axis current, and the difference between the two is calculated to obtain the q-axis current deviation, i.e.: ; in, For the first q-axis current deviation per switching cycle For the first The q-axis current command for each switching cycle (from the speed loop proportional-integral control output in step 22). For the first The actual q-axis current for each switching cycle; all parameters are in amperes (A). A positive q-axis current deviation indicates that the actual torque current is less than the commanded value, requiring an increase in the q-axis voltage to make the actual current rise and track the command. A negative q-axis current deviation indicates that the actual torque current exceeds the commanded value, requiring a decrease in the q-axis voltage to make the actual current fall back. This step yields the d-axis current deviation and the q-axis current deviation.

[0051] Step 32: Based on the d-axis current deviation, perform d-axis current loop proportional-integral adjustment to obtain the d-axis reference voltage vector.

[0052] Specifically, in this step, the execution process of the d-axis current loop proportional-integral adjustment is as follows: Based on the d-axis current deviation output in step 31, d-axis current loop proportional-integral adjustment is performed. The d-axis current loop proportional-integral adjustment includes two parts: a proportional calculation channel and an integral calculation channel.

[0053] The d-axis current deviation of the current switching cycle is multiplied by a preset d-axis current loop proportional gain coefficient to obtain the d-axis proportional adjustment component. The value of the d-axis current loop proportional gain coefficient is determined in advance based on the d-axis inductance parameters, stator resistance parameters, and desired bandwidth of the permanent magnet synchronous motor. Its dimension is volts per ampere (V / A), and its physical meaning is the amount of d-axis voltage adjustment generated per unit d-axis current deviation.

[0054] An integral calculation is performed on the d-axis current deviation. Specifically, within the current switching cycle, the d-axis current deviation of the current switching cycle is multiplied by a preset d-axis current loop integral gain coefficient, and then multiplied by the time step of the switching cycle to obtain the d-axis integral increment of the current switching cycle. The d-axis integral increment is added to the cumulative d-axis integral value of the previous switching cycle to obtain the d-axis integral adjustment component of the current switching cycle. The value of the d-axis current loop integral gain coefficient is predetermined based on the steady-state accuracy requirements of the current loop and the integral saturation constraint. Its dimension is volts per ampere per second (V / (A⋅s)), and its physical meaning is the d-axis voltage integral adjustment generated per unit d-axis current deviation per unit time.

[0055] During the integral accumulation process, anti-integral saturation processing is performed on the d-axis integral adjustment component, that is: the absolute value of the d-axis integral adjustment component is compared with a preset d-axis integral limiting threshold; if it exceeds the d-axis integral limiting threshold, the d-axis integral adjustment component is clamped to a limiting value with the same sign as the d-axis integral limiting threshold, and the integral accumulation operation of the current switching cycle is paused until the d-axis current deviation falls back to the normal adjustment range before the integral accumulation is resumed.

[0056] Based on this, the d-axis proportional adjustment component is added to the d-axis integral adjustment component after anti-integral saturation processing to obtain the d-axis reference voltage vector.

[0057] After obtaining the d-axis reference voltage vector, its amplitude is limited. Specifically, the absolute value of the d-axis reference voltage vector is compared with a preset d-axis voltage limiting threshold. If the absolute value exceeds the threshold, the d-axis reference voltage vector is clamped to a limiting value with the same sign as the threshold. If the absolute value does not exceed the threshold, the original value is maintained. The d-axis voltage limiting threshold is preset based on the DC bus voltage of the GaN device three-phase full-bridge inverter and the rated voltage constraint of the permanent magnet synchronous motor, and is expressed in volts (V). This step yields the d-axis reference voltage vector.

[0058] This step introduces both proportional and integral regulation into the d-axis current loop, enabling the d-axis current to respond quickly to changes in the current command (zero current) and eliminate steady-state current error. The anti-integral saturation treatment prevents the regulator from desaturating and overshooting due to continuous integral accumulation when the d-axis voltage output reaches the limiting boundary, thus ensuring the speed and stability of d-axis current regulation.

[0059] Step 33: Based on the q-axis current deviation, perform q-axis current loop proportional-integral adjustment to obtain the q-axis reference voltage vector.

[0060] Specifically, in this step, the execution process of the q-axis current loop proportional-integral adjustment is as follows: Based on the q-axis current deviation output in step 31, q-axis current loop proportional-integral adjustment is performed. The q-axis current loop proportional-integral adjustment comprises two parts: a proportional calculation channel and an integral calculation channel.

[0061] The q-axis current deviation of the current switching cycle is multiplied by a preset q-axis current loop proportional gain coefficient to obtain the q-axis proportional adjustment component. The value of the q-axis current loop proportional gain coefficient is determined in advance based on the q-axis inductance parameters, stator resistance parameters, and desired bandwidth of the permanent magnet synchronous motor. Its dimension is volts per ampere (V / A), and its physical meaning is the amount of q-axis voltage adjustment generated per unit q-axis current deviation.

[0062] An integral calculation is performed on the q-axis current deviation. Specifically, within the current switching cycle, the q-axis current deviation of the current switching cycle is multiplied by a preset q-axis current loop integral gain coefficient, and then multiplied by the time step of the switching cycle to obtain the q-axis integral increment of the current switching cycle. This q-axis integral increment is then added to the accumulated q-axis integral value of the previous switching cycle to obtain the q-axis integral adjustment component of the current switching cycle. The value of the q-axis current loop integral gain coefficient is predetermined based on the steady-state accuracy requirements of the current loop and the integral saturation constraint. Its dimension is volts per ampere per second, and its physical meaning is the q-axis voltage integral adjustment generated per unit q-axis current deviation per unit time.

[0063] During the integral accumulation process, anti-integral saturation processing is performed on the q-axis integral adjustment component, that is: the absolute value of the q-axis integral adjustment component is compared with a preset q-axis integral limiting threshold; if it exceeds the q-axis integral limiting threshold, the q-axis integral adjustment component is clamped to a limiting value with the same sign as the q-axis integral limiting threshold, and the integral accumulation operation of the current switching cycle is paused until the q-axis current deviation falls back to the normal adjustment range before the integral accumulation is resumed.

[0064] Based on this, the q-axis proportional adjustment component is added to the q-axis integral adjustment component after anti-integral saturation processing to obtain the q-axis reference voltage vector.

[0065] After obtaining the q-axis reference voltage vector, its amplitude is limited. Specifically, the absolute value of the q-axis reference voltage vector is compared with a preset q-axis voltage limiting threshold. If it exceeds the threshold, the q-axis reference voltage vector is clamped to a limiting value with the same sign as the threshold. If it does not exceed the threshold, the original value remains unchanged. The q-axis voltage limiting threshold is preset based on the DC bus voltage of the GaN device three-phase full-bridge inverter and the rated voltage constraint of the permanent magnet synchronous motor, and is measured in volts (V). This step yields the q-axis reference voltage vector.

[0066] This step introduces both proportional and integral regulation into the q-axis current loop, enabling the q-axis current to quickly respond to changes in the q-axis current command output from the speed loop and eliminating steady-state current errors. The anti-integral saturation treatment prevents the regulator from desaturating and overshooting due to continuous integral accumulation when the q-axis voltage output reaches the limiting boundary, ensuring the speed and stability of q-axis current regulation, and thus ensuring that the electromagnetic torque can quickly track changes in load demand.

[0067] Step 34: Based on the d-axis reference voltage vector and the q-axis reference voltage vector, and the electrical angle calculated based on the rotor mechanical angular position, the α-axis reference voltage component and the β-axis reference voltage component in the α-β stationary coordinate system are obtained through inverse Park transformation, and used as the reference voltage vector in the stationary coordinate system.

[0068] Specifically, in this step, the inverse Parker transformation is performed as follows: Based on the rotor mechanical angular position output in step 11, the electrical angle is calculated. The calculation method is as follows: multiply the rotor mechanical angular position by the number of pole pairs of the permanent magnet synchronous motor, and perform a reduction process with a period of 360 degrees of electrical angle to ensure that the electrical angle is always within the range of 0 degrees to 360 degrees of electrical angle, thus obtaining the electrical angle corresponding to the current rotor position.

[0069] Based on the d-axis reference voltage vector output in step 32, the q-axis reference voltage vector output in step 33, and the electrical angle, the voltage vector in the dq rotating coordinate system is rotated and transformed to the α-β stationary coordinate system. During the transformation, the product of the d-axis reference voltage vector and the cosine of the electrical angle is subtracted from the product of the q-axis reference voltage vector and the sine of the electrical angle to obtain the α-axis reference voltage component. Simultaneously, the product of the d-axis reference voltage vector and the sine of the electrical angle is added to the product of the q-axis reference voltage vector and the cosine of the electrical angle to obtain the β-axis reference voltage component. This inverse Park transformation maintains the principle of amplitude invariance, meaning the amplitude of the synthesized voltage vector before and after the transformation is equal. The α-axis reference voltage component and the β-axis reference voltage component together constitute the reference voltage vector in the stationary coordinate system.

[0070] This step yields a reference voltage vector in the stationary coordinate system, which serves as the input reference for the MPC module to perform voltage space vector optimization in the subsequent step 4.

[0071] This step, through inverse Parker transformation, converts the d-axis and q-axis reference voltage vectors output by the current loop regulator in a rotating coordinate system to a stationary coordinate system fixed to the stator windings. This ensures that the reference voltage vectors are in the same coordinate system as the basic voltage space vector of the GaN device three-phase full-bridge inverter, providing a unified coordinate reference for the MPC module to directly select the optimal voltage space vector within the current switching cycle. This guarantees seamless integration between the FOC current loop output and MPC online optimization. Since the FOC-MPC fusion architecture of this invention no longer undergoes traditional SVPWM or SPWM modulation after inverse Parker transformation, but instead directly sends the reference voltage vector in the stationary coordinate system to the MPC module for voltage vector optimization, there is no modulation delay between the current loop regulator outputting the reference voltage and the inverter applying the drive voltage. Therefore, the current loop voltage command can be applied to the motor stator windings without delay within the current switching cycle.

[0072] In a preferred embodiment of the present invention, step 4 includes: Step 41: Based on the reference voltage vector in the stationary coordinate system, within the current switching cycle of the GaN device three-phase full-bridge inverter, using the stator current sampling value at the current moment, calculate the predicted current value corresponding to each basic voltage space vector of the GaN device three-phase full-bridge inverter at the next sampling moment, based on the discretized current prediction model of the permanent magnet synchronous motor in the stationary coordinate system.

[0073] Step 411, based on the output of step 12 Axis current components and The shaft current component is used to obtain the stator current at the start of the current switching cycle. - The sampled values ​​in the stationary coordinate system are denoted as follows: Current current sampling value of shaft and The current sampled value of the axis is used as the initial state variable of the current prediction model. The sampling trigger time at the start of the current switching cycle is aligned with the start edge of the current switching cycle of the GaN device three-phase full-bridge inverter.

[0074] Based on this, in step 412, the electrical angular velocity corresponding to the current switching cycle is calculated according to the actual rotor speed output in step 14 and the number of pole pairs of the permanent magnet synchronous motor. The electrical angular velocity is the electrical rotational angular frequency in radians per second obtained by multiplying the actual rotor speed (in revolutions per minute) by the number of pole pairs and then by a conversion factor of 120 radians per 60 seconds. Its value is directly used for the calculation of the back electromotive force term in the subsequent discretized current prediction model.

[0075] Based on this, step 413 involves obtaining the stator resistance parameters, stator inductance parameters, and permanent magnet flux linkage parameters of the permanent magnet synchronous motor, which were previously identified offline. The stator resistance parameters are obtained through DC injection experiments, and their physical meaning is the equivalent resistance value of each phase winding of the motor stator, expressed in ohms. The stator inductance parameters are obtained through AC injection experiments. For surface-mounted permanent magnet synchronous motors, the d-axis inductance and q-axis inductance values ​​are equal and are collectively referred to as stator inductance parameters. Their physical meaning is the equivalent synchronous inductance value of each phase winding of the motor stator, expressed in Henry (H). The permanent magnet flux linkage parameters are obtained through back electromotive force measurement. Their physical meaning is the flux linkage amplitude of the rotor permanent magnet in each phase winding of the stator, expressed in Weber (Wb). These motor parameters are stored in the parameter register of the servo driver and are called by the current prediction model in each switching cycle.

[0076] Based on this, step 414, according to the information obtained in step 411 Current current sampling value of shaft and The current current sampling value of the shaft, the electric angular velocity calculated in step 412, the stator resistance parameters, stator inductance parameters, and permanent magnet flux linkage parameters obtained in step 413, and the preset switching cycle time step (i.e., the reciprocal of the preset switching frequency of the GaN device three-phase full-bridge inverter) are used to construct the permanent magnet synchronous motor in... - Discretized current prediction model in stationary coordinate system.

[0077] In this step, the construction logic of the discretized current prediction model is as follows: Establish permanent magnet synchronous motor in - The continuous-time state equation in a stationary coordinate system. This state equation is based on... Axis current components and The shaft current component is a state variable, with Axis voltage components and The d-axis voltage component is the input variable, and the electric angular velocity and permanent magnet flux linkage constitute the back electromotive force coupling term. For a surface-mounted permanent magnet synchronous motor, the d-axis inductance is equal to the q-axis inductance, and they are collectively denoted as the stator inductance parameter. The expression for the continuous-time state equation is: ; ; in, for Axial current component, in amperes (A); for Axial current component, in amperes (A); for Axial voltage component, in volts (V); for Axial voltage component, in volts (V); These are stator resistance parameters, in ohms (Ω). ); These are stator inductance parameters, expressed in Henry (H). ω is the electric angular velocity, measured in radians per second (rad / s). is the flux linkage parameter of the permanent magnet, in Weber (Wb). The value is an electrical angle (i.e., the result of multiplying the rotor mechanical angular position output in step 11 by the number of magnetic pole pairs and then performing a 360-degree reduction), in radians (rad). Time is measured in seconds (s).

[0078] Based on this, the first-order forward Euler discretization method is applied to the current derivative term in the above continuous-time state equation, approximating the current derivative as the ratio of the current change between the current sampling time and the next sampling time to the time step of the switching cycle, i.e.: ; in, For the first The current sample value at each sampling time (the start time of the current switching cycle). For the first The predicted current value at each sampling time (the start time of the next switching cycle). The switching cycle time step is in seconds (s) and its value is equal to the reciprocal of the preset switching frequency of the GaN device three-phase full-bridge inverter.

[0079] Substituting the first-order forward Euler approximation into the continuous-time state equations and rearranging, we obtain the state equations for the permanent magnet synchronous motor. - Discretized current prediction model in stationary coordinate system: ; ; in, For the information obtained in step 411 Current current sample value of the shaft, in amperes (A); For the information obtained in step 411 Current current sample value of the shaft, in amperes (A); In the first Each sampling time The axis predicted current component is in amperes (A). In the first Each sampling time The axis predicted current component is in amperes (A). The applied during the current switching cycle Axial voltage component, in volts (V); The applied during the current switching cycle Axial voltage component, in volts (V); The current electric angular velocity calculated in step 412 is expressed in radians per second (rad / s). The current electrical angle is expressed in radians (rad). The switching cycle time step is in seconds (s). These are stator resistance parameters, in ohms (Ω). ); These are stator inductance parameters, expressed in Henry (H). This is the flux linkage parameter of the permanent magnet, in Weber (Wb).

[0080] The recursive logic of this discretized current prediction model is as follows: the predicted current value at the next moment equals the current sample value at the current moment multiplied by the natural current decay coefficient, plus the current increment generated by the applied voltage vector at the current moment, plus (or minus) the current component generated by the kinetic back electromotive force induced by the rotation of the rotor permanent magnet in the stator winding. Here, the natural current decay coefficient is jointly determined by the stator resistance parameters, stator inductance parameters, and the switching cycle time step, characterizing the proportion of current decay caused by the stator resistance voltage drop within one switching cycle; the current increment generated by the voltage is jointly determined by the applied voltage amplitude, stator inductance parameters, and the switching cycle time step; the back electromotive force current component is jointly determined by the electric angular velocity, permanent magnet flux linkage parameters, stator inductance parameters, and the switching cycle time step, and is further determined by... shaft and The axial directions are allocated according to the sine and cosine functions of the electrical angle, respectively.

[0081] Based on this, in step 415, according to the discretized current prediction model constructed in step 414, each basic voltage space vector of the GaN device three-phase full-bridge inverter is... - The corresponding coordinate system in the stationary coordinate system Axis voltage components and The axis voltage components are treated one by one as and Substitute the voltage input term into the discretized current prediction model.

[0082] Specifically, in this step, the total number of basic voltage space vectors in the GaN device three-phase full-bridge inverter is eight, including six non-zero voltage space vectors and two zero voltage space vectors. Each basic voltage space vector corresponds to a specific... Axis voltage components and Combination of axis voltage components. Among them, the first... The basic voltage space vectors ( ) corresponding The axis voltage component is denoted as , The axis voltage component is denoted as Both are uniquely determined by the DC bus voltage of the GaN device three-phase full-bridge inverter and the switching state combination corresponding to the basic voltage space vector, and their calculation process is a well-known basic content in this field.

[0083] Based on this, for each basic voltage space vector, the values ​​obtained in step 411 are... Current current sampling value of shaft and Current current sampling value of shaft Substitute the initial state variables into the model, and calculate the current electric angular velocity in step 412. Substituting the back electromotive force term into the model, the current electric angle Substituting the sine and cosine function terms into the model, and setting the preset switching cycle time step... and the stator resistance parameters obtained in step 413 Stator inductance parameters and permanent magnet flux linkage parameters Substituting the coefficients into the model, calculate the value of each basic voltage space vector at the next sampling time after one switching cycle. Axis Predicted Current Components and Axis Predicted Current Components Together, they constitute the first The predicted current value corresponding to each basic voltage space vector.

[0084] For the The basic voltage space vectors, their corresponding Axis predicted current components and The formulas for calculating the predicted current components of the shaft are as follows: ; ; Among them, subscript Indicates the first The index number of each basic voltage space vector. These correspond to the eight basic voltage space vectors, respectively. For the first The basic voltage space vectors corresponding to The axis predicted current component is in amperes (A). For the first The basic voltage space vectors corresponding to The axis predicted current component is in amperes (A). For the first The basic voltage space vectors corresponding to Axial voltage component, in volts (V); For the first The basic voltage space vectors corresponding to The axis voltage component is in volts (V); the meanings and units of the remaining parameters are the same as those described in step 414.

[0085] This step yields eight sets of predicted current values ​​that correspond one-to-one with the eight fundamental voltage space vectors of a GaN device three-phase full-bridge inverter. Each set of predicted current values ​​contains... Axis predicted current components and The axis predicts the current component, with a total of sixteen scalar values.

[0086] This step involves traversing and predicting all eight basic voltage space vectors within each switching cycle. Using a discretized current prediction model, it accurately estimates the current response of each voltage vector after one switching cycle, providing complete predicted current information for subsequent steps, including cost function calculation and optimal voltage vector selection. Since the discretized current prediction model is directly based on the physical state equations of the permanent magnet synchronous motor in the stationary coordinate system through first-order forward Euler discretization, its prediction accuracy is unaffected by the linearization approximation error of the cross-coupling terms in the rotating coordinate system. Furthermore, the model is reinitialized within each switching cycle using the current actual sampled current as the initial state, eliminating the error accumulation problem in open-loop prediction and ensuring high accuracy in short-term predictions. This provides a reliable foundation for the online optimization of the MPC module.

[0087] Step 42: Based on the current tracking error between the predicted current value corresponding to each basic voltage space vector and the reference current value corresponding to the reference voltage vector in the stationary coordinate system, construct a cost function with the goal of minimizing the current tracking error.

[0088] Step 421: Obtain the reference current value in the stationary coordinate system. Based on the output of step 22... Shaft current command and Shaft current command (where (The shaft current command is zero), and the current electrical angle is calculated based on the rotor mechanical angular position output in step 11. (The calculation method is the same as described in step 414), and after inverse Parker transformation, we obtain... - stationary coordinate system Shaft reference current components and Shaft reference current component.

[0089] In this step, the execution process of the inverse Park transform is as follows: Because The shaft current command is zero. Shaft reference current components and The calculation of the shaft reference current component is simplified to only by The shaft current command is combined with the sine and cosine functions of the electrical angle, and its expression is as follows: ; ; in, In the first Each sampling time Shaft reference current component, in amperes (A); In the first Each sampling time Shaft reference current component, in amperes (A); The output of step 22 Shaft current command, in amperes (A); The current electrical angle is expressed in radians (rad).

[0090] Shaft reference current components and The axis reference current components together constitute the reference current value in the stationary coordinate system. This reference current value characterizes the stator current vector target that the current loop expects to achieve in the current switching cycle, and it maintains consistency with the reference voltage vector in the stationary coordinate system output in step 34 in the control target direction.

[0091] Based on this, in step 422, according to each basic voltage space vector output in step 41, Axis Predicted Current Components and Axis Predicted Current Components and the output of step 421 Shaft reference current components and Shaft reference current components For each basic voltage space vector, calculate separately Shaft predicted current components and The difference between the shaft reference current components, and Shaft predicted current components and The difference between the axis reference current components.

[0092] For the The basic voltage space vector, its Axis current tracking error and The formula for calculating the shaft current tracking error is: ; ; in, For the first The basic voltage space vectors corresponding to Shaft current tracking error, in amperes (A); For the first The basic voltage space vectors corresponding to Shaft current tracking error, in amperes (A).

[0093] when When the axis current tracking error is positive, it indicates that under the action of the basic voltage space vector, the next sampling time... The predicted current component of the shaft will be greater than the reference current component, and the actual current will tend to overshoot; when When the axis current tracking error is negative, it indicates that the predicted current component will be less than the reference current component, and the actual current has an under-adjustment trend. The positive and negative meanings of axis current tracking error are similar.

[0094] Based on this, in step 423, according to each basic voltage space vector output in step 422... Axis current tracking error and The cost function is constructed based on the current tracking error. The cost function aims to minimize the current tracking error, and its construction logic is as follows: [The text abruptly ends here, so the translation stops.] The square of the axis current tracking error and The squares of the axis current tracking errors are summed, and this sum is taken as the cost function value corresponding to the basic voltage space vector.

[0095] For the The expression for the cost function of the basic voltage space vectors is as follows: ; Substituting the definition of current tracking error from step 422, the cost function can be further written as: ; in, For the first The cost function value corresponding to each basic voltage space vector, in amperes squared (A). 2 ); The output of step 41 The basic voltage space vectors corresponding to The axis predicted current component is in amperes (A). The output of step 41 The basic voltage space vectors corresponding to The axis predicted current component is in amperes (A). The output of step 421 Shaft reference current component, in amperes (A); The output of step 421 Shaft reference current component, in amperes (A).

[0096] In this step, the cost function value The physical meaning is: in - In a stationary coordinate system, the first The cost function is the square of the Euclidean distance between the predicted current vector and the reference current vector corresponding to each basic voltage space vector. A smaller cost function value indicates that, under the influence of that basic voltage space vector, the predicted current vector at the next sampling time will be... shaft and The closer the combined current vector is to the reference current vector in the two orthogonal directions of the axis, the higher the current tracking accuracy; conversely, the larger the cost function value, the greater the current tracking deviation.

[0097] This step yields eight cost function values ​​that correspond one-to-one with the eight fundamental voltage space vectors of the GaN device three-phase full-bridge inverter, denoted as follows: , , , , , , , .

[0098] This step will... shaft and The current tracking errors in the two orthogonal directions of the axis are uniformly incorporated into a scalarized cost function for comprehensive evaluation. This allows the MPC module to quantitatively compare the current tracking performance of each basic voltage space vector in a stationary coordinate system. The cost function directly uses the quadratic form of the deviation between the predicted current and the reference current as the evaluation index, which is completely consistent with the control objective of the current loop, ensuring that the actual current tracks the current command quickly and accurately. This guarantees the inherent consistency between the MPC optimization results and the FOC current loop control objective. Furthermore, the calculation process of the cost function involves only subtraction, multiplication, and addition operations. The cost function calculation for each basic voltage space vector requires only two subtractions, two multiplications, and one addition. For the eight basic voltage space vectors, this results in a total of sixteen subtractions, sixteen multiplications, and eight additions, resulting in extremely low computational complexity. This allows the calculation of all eight cost functions to be completed within the extremely short switching cycle corresponding to the high switching frequency of GaN devices, meeting real-time control requirements.

[0099] Step 43: Select the final voltage space vector that minimizes the cost function from all the basic voltage space vectors of the GaN device three-phase full-bridge inverter.

[0100] Step 431: Convert the eight cost function values ​​corresponding to the eight basic voltage space vectors output in Step 42. , , , , , , , The minimum value is determined by comparing each pair of values ​​one by one and using a successive comparison optimization method.

[0101] The specific execution process of successive comparison and optimization is as follows: The first basic voltage space vector ( The cost function value As a candidate for the current minimum value, let it be denoted as Simultaneously, the first basic voltage space vector is selected as a candidate for the current optimal voltage space vector, denoted as... The second fundamental voltage space vector ( The cost function value Compared with the current minimum candidate Comparison: If Less than Then Update to the new current minimum candidate Simultaneously, the second basic voltage space vector is updated to a new current optimal voltage space vector candidate. ;like Greater than or equal to Then keep the current minimum candidate and current optimal voltage space vector candidate Unchanged. According to... In the order of the remaining six basic voltage space vectors, the cost function values ​​corresponding to each vector are sequentially assigned. Compared with the current minimum candidate Perform the same comparison and update operations as described above.

[0102] After completing the traversal and comparison of all eight basic voltage space vectors, the final candidate for the current minimum value is determined. This is the global minimum among all eight cost function values. The fundamental voltage space vector corresponding to this global minimum is... That is, the voltage space vector that minimizes the cost function among all eight basic voltage space vectors.

[0103] Based on this, step 432: according to the global minimum cost function value determined in step 431 The basic voltage space vector corresponding to the minimum cost function value It is determined as the final voltage space vector acting on the GaN device three-phase full-bridge inverter during the current switching cycle.

[0104] If two or more basic voltage space vectors have the same cost function value and are both global minimum values, the basic voltage space vector with the smaller voltage amplitude is selected as the final voltage space vector. If the voltage amplitudes are still the same, one of the zero voltage space vectors is selected as the final voltage space vector to reduce switching losses.

[0105] This step yields the final voltage space vector, which minimizes the tracking error between the predicted current vector and the reference current vector at the next sampling time among all eight basic voltage space vectors.

[0106] This step involves a successive comparison and optimization process to traverse and search through the eight basic voltage space vectors, ensuring that the final selected voltage space vector is the globally optimal solution, rather than a local approximation or suboptimal solution. Compared to the traditional SVPWM modulation stage, which synthesizes a reference voltage vector with arbitrary amplitude and phase using the volt-second balance principle, introducing a computational delay of one to two cycles in digital implementation, this step transforms the voltage vector selection problem into a discrete optimization problem within a finite set. It directly selects the one that best matches the current tracking target from the eight basic voltage space vectors that the inverter can actually output, eliminating the delay in the modulation stage. Furthermore, since there are only eight basic voltage space vectors, the computational load of the traversal search is fixed and controllable. The number of comparison operations in each optimization is constant at seven, and the computation time remains consistent regardless of changes in the amplitude and phase of the reference voltage vector. This is beneficial for achieving deterministic real-time control within each switching cycle at the high switching frequency of GaN devices.

[0107] Step 44: Based on the final voltage space vector, determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter during the current switching cycle.

[0108] Step 441: Obtain the pre-established mapping table of voltage space vectors and switching states. This mapping table is stored in the program memory of the servo driver in the form of a look-up table, defining a one-to-one correspondence between all eight basic voltage space vectors of the GaN device three-phase full-bridge inverter and the switching states of the six GaN switches in the three bridge arms of the inverter.

[0109] In this mapping table, each basic voltage space vector is identified by a three-bit binary code. From left to right, this three-bit binary code represents the switching states of the upper transistor in phase A, phase B, and phase C bridge arms, respectively. A binary value of "1" indicates that the upper transistor of the corresponding bridge arm is in the ON state and the lower transistor is in the OFF state, while a binary value of "0" indicates that the upper transistor of the corresponding bridge arm is in the OFF state and the lower transistor is in the ON state. The six non-zero voltage space vectors correspond to six switching state combinations: (1,0,0), (1,1,0), (0,1,0), (0,1,1), (0,0,1), and (1,0,1). The two zero voltage space vectors correspond to two switching state combinations: (0,0,0) and (1,1,1).

[0110] Based on this, in step 442, according to the final voltage space vector output in step 432, the entry that matches the final voltage space vector is searched in the mapping table described in step 441, and the three-bit binary code corresponding to the entry is read.

[0111] The status of the upper pipe switch of phase A bridge arm is determined based on the first bit of the three-bit binary code; the status of the upper pipe switch of phase B bridge arm is determined based on the second bit; and the status of the upper pipe switch of phase C bridge arm is determined based on the third bit.

[0112] Based on this, in step 443, according to the switch status of the upper tube of phase A bridge arm, the switch status of the upper tube of phase B bridge arm, and the switch status of the upper tube of phase C bridge arm output in step 442, and combined with the basic operating rules of complementary conduction between the upper and lower tubes of each bridge arm of the GaN device three-phase full-bridge inverter, the switch status of the lower tube of phase A bridge arm, the switch status of the lower tube of phase B bridge arm, and the switch status of the lower tube of phase C bridge arm are determined.

[0113] The basic operating rule of the complementary conduction is as follows: In the same bridge arm, the upper and lower tubes must not be turned on at the same time to avoid bridge arm short circuit faults. That is, the switch state of the lower tube of each bridge arm is the logical inverse of the switch state of the upper tube of that bridge arm. When the switch state of the upper tube is "1" (on), the switch state of the lower tube is "0" (off); when the switch state of the upper tube is "0" (off), the switch state of the lower tube is "1" (on).

[0114] During the transitions from the on state to the off state and vice versa, a preset dead time interval is inserted between the drive signals of the upper and lower transistors. Within this dead time interval, both the upper and lower transistors of the same bridge arm are in the off state to prevent bridge arm shoot-through faults caused by the slight time overlap between the turn-on and turn-off actions of the GaN switches. The value of the dead time interval is preset based on the turn-on and turn-off delay characteristics of the GaN switches, and is on the order of nanoseconds.

[0115] At this point, the on / off states of the six GaN switching transistors—the upper and lower transistors of phase A bridge arm, the upper and lower transistors of phase B bridge arm, the upper and lower transistors of phase C bridge arm, and the lower transistor of phase C bridge arm—are determined during the current switching cycle, and together they constitute the final switching state combination during the current switching cycle.

[0116] This step yields the final switching state combination, which will be directly used to drive each GaN switch in the three-phase full-bridge inverter during the current switching cycle. Subsequently, this final switching state combination will serve as the input signal for the gate drive circuit in step 5, controlling each GaN switch to perform the corresponding turn-on or turn-off action.

[0117] This step achieves instantaneous conversion from the final voltage space vector to the final switching state combination through a pre-established lookup table. The conversion process eliminates the intermediate processing steps required by traditional SVPWM modulation, such as sector judgment, vector action time calculation, and PWM comparison value generation. The conversion delay depends only on the hardware execution time of the lookup operation and logic inversion operation, and can be completed within a very small proportion of a switching cycle at the high switching frequency of GaN devices. Therefore, the entire decision chain, from the current loop regulator outputting the reference voltage vector (step 34) to the MPC module selecting the final voltage space vector (step 43) and then determining the final switching state combination (this step), is completed within the current switching cycle. This achieves zero-delay conversion of the current loop voltage command into the specific actions of the inverter switching transistors within the current switching cycle, eliminating the one- to two-step calculation delay and modulation delay introduced by the modulation stage in traditional FOC control strategies. Consequently, the effective bandwidth of the current loop is significantly improved.

[0118] In a preferred embodiment of the present invention, step 5 includes: Step 51: Based on the final switching state combination, generate the gate drive logic signals corresponding to each GaN switch in the three-phase full-bridge inverter of the GaN device within the current switching cycle; specifically including: Step 511: Based on the final switching state combination output in step 44, obtain the target switching state of each of the six GaN switches in the three-phase full-bridge inverter with GaN devices during the current switching cycle. The final switching state combination includes the switching states of the upper and lower switches of phase A bridge arms, the upper and lower switches of phase B bridge arms, the upper and lower switches of phase B bridge arms, the upper and lower switches of phase C bridge arms, and the value of each switching state is either on or off.

[0119] Based on this, in step 512, the target switching states of each GaN switch obtained in step 511 are converted into corresponding gate drive logic signals. The gate drive logic signals are digital logic level signals. For GaN switches whose target switching state is on, the corresponding gate drive logic signal is an on logic level; for GaN switches whose target switching state is off, the corresponding gate drive logic signal is an off logic level. In this step, the gate drive logic signals are low-voltage digital signals, and their voltage amplitude is within the power supply voltage range of the digital logic circuit. This is insufficient to directly drive the gate of the GaN switch; it needs to be level-converted and power-amplified by the subsequent gate drive circuit before it can be applied to the gate of the GaN switch.

[0120] Based on this, in step 513, according to the gate drive logic signals corresponding to each GaN switch generated in step 512, and combined with the basic operating rule of complementary conduction between the upper and lower switches in the same bridge arm, dead-time insertion processing is performed on the gate drive logic signals. The specific execution process of the dead-time insertion processing is as follows: For each bridge arm, when it is detected that the gate drive logic signals of the upper and lower GaN switches are about to undergo a state transition (i.e., one signal transitions from an on logic level to an off logic level and the other from an off logic level to an on logic level), a preset delay is inserted into the gate drive logic signal of the GaN switch about to be turned on. This causes the turn-on time of the GaN switch to lag behind the turn-off time of the other GaN switch in the same bridge arm by a dead time interval. Therefore, during this dead time interval, the gate drive logic signals of both the upper and lower switches in the same bridge arm are at an off logic level. The value of the dead time interval is preset based on the turn-on and turn-off delay characteristics of the GaN switches, typically ranging from several nanoseconds to tens of nanoseconds.

[0121] Based on this, in step 514, according to the gate drive logic signal corresponding to each GaN switch after the dead-time insertion process in step 513, and combined with the resonant state of the resonant network set between the DC bus of the three-phase full-bridge inverter of the GaN device and the midpoint of each bridge arm in the current switching cycle, soft switching timing adjustment is performed on the flip-off time of the gate drive logic signal. The soft-switching timing adjustment process is as follows: For a GaN switch about to perform a turn-on operation, based on the resonant zero-crossing detection signal of the voltage across the resonant capacitor connected in parallel with the drain-source of the GaN switch in the resonant network, the moment when the gate drive logic signal is flipped from the turn-off logic level to the turn-on logic level is precisely aligned with the zero-voltage moment when the drain-source voltage of the GaN switch naturally resonates to zero, so as to achieve zero-voltage turn-on; For a GaN switch about to perform a turn-off operation, based on the resonant zero-crossing detection signal of the current in the resonant inductor connected in series with the GaN switch in the resonant network, the moment when the gate drive logic signal is flipped from the turn-on logic level to the turn-off logic level is precisely aligned with the zero-current moment when the drain-source current of the GaN switch naturally resonates to zero, so as to achieve zero-current turn-off.

[0122] In this step, after soft-switching timing adjustment, the flipping time of the gate drive logic signal of each GaN switch is no longer simply aligned with the fixed phase of the switching cycle, but is dynamically determined according to the natural resonance process of the resonant network, so that each switching action occurs near the natural zero point of voltage or current, thereby minimizing switching losses.

[0123] This step yields six gate drive logic signals corresponding to the six GaN switches in the GaN device three-phase full-bridge inverter within the current switching cycle, after dead-time insertion and soft-switching timing adjustment.

[0124] Step 52: According to the gate drive logic signal, for the GaN switch about to be turned on, an active Miller clamping circuit integrated inside the SiP package module and arranged adjacent to the gate pad of the GaN switch clamps the gate voltage at a preset negative voltage level during the gate turn-off period to suppress the gate voltage erroneous turn-on caused by the Miller effect due to the drain-source voltage change rate; specifically including: Step 521: Based on the six gate drive logic signals output in step 51, identify the GaN switch corresponding to the gate drive logic signal that is at the off logic level in the current switching cycle, and determine these GaN switch as the clamping objects of the active Miller clamping circuit.

[0125] Based on this, in step 522, for each clamped GaN switch determined in step 521, an active Miller clamping circuit is obtained, integrated within the SiP package module and arranged adjacent to the gate pad of the GaN switch. The active Miller clamping circuit includes a clamping switch and a clamping voltage reference source. The drain of the clamping switch is connected to the gate of the GaN switch, and the source of the clamping switch is connected to the output of the clamping voltage reference source, which provides a preset negative voltage level. The value of the preset negative voltage level is preset based on the safety margin of the gate threshold voltage of the GaN switch and the Miller capacitance coupling voltage. For enhancement-mode GaN high electron mobility transistors, the typical gate threshold voltage is between 1.0V and 2.0V, while the typical preset negative voltage level is between -3.0V and -5.0V. This ensures that even under the worst drain-source voltage change conditions, the gate voltage remains far below the gate threshold voltage after the voltage offset coupled to the gate by the Miller capacitance is superimposed with the preset negative voltage level, thereby preventing false gate turn-on due to the Miller effect.

[0126] Based on this, in step 523, according to the gate drive logic signal corresponding to the GaN switch output in step 51, when the gate drive logic signal is at the off logic level, the clamping switch in the active Miller clamping circuit is turned on, and the gate of the GaN switch is connected to the output terminal of the clamping voltage reference source through a low-impedance path, so that the gate voltage of the GaN switch is actively clamped at a preset negative voltage level. When the gate drive logic signal flips from the off logic level to the on logic level, the clamping switch in the active Miller clamping circuit is turned off, releasing the clamping of the gate of the GaN switch, so that the subsequent negative voltage direct drive gate drive circuit can normally provide a positive drive voltage pulse to the gate.

[0127] In this step, the active Miller clamping circuit keeps the clamping switch on during the turn-off period, fixing the gate voltage at a preset negative voltage level through a low-impedance path. Therefore, when the drain-source voltage of the GaN switch changes rapidly during the inverter commutation process, i.e., when a high drain-source voltage change rate is generated, the displacement current coupled to the gate through the Miller capacitor is absorbed by the low-impedance path of the clamping switch. The gate voltage will not rise positively due to the Miller effect, thereby effectively suppressing the risk of bridge arm shoot-through faults.

[0128] This step completes the active Miller clamping of the gate voltage of all GaN switches that are in the off state during the current switching cycle. The gate voltage of each clamped GaN switch is stably maintained at a preset negative voltage level, providing a stable negative voltage baseline for the turn-off operation of the negative voltage direct drive gate drive circuit in subsequent steps.

[0129] Step 53: According to the gate drive logic signal, a negative voltage direct drive gate drive circuit cascaded with the active Miller clamp circuit provides a positive drive voltage pulse to the gate of the GaN switch during the gate turn-on period to turn it on, and provides a negative turn-off voltage pulse to the gate of the GaN switch during the gate turn-off period to turn it off; specifically including: Step 531: Based on the six gate drive logic signals, determine the drive type of each GaN switch in the current switching cycle. When the gate drive logic signal corresponding to the GaN switch is at the turn-on logic level, determine its drive type as turn-on drive; when the gate drive logic signal corresponding to the GaN switch is at the turn-off logic level, determine its drive type as turn-off drive.

[0130] Based on this, in step 532, for the GaN switch transistor whose drive type is determined to be turn-on drive in step 531, a negative voltage direct-drive gate drive circuit connected to the gate of the GaN switch transistor is obtained. The negative voltage direct-drive gate drive circuit includes a positive voltage drive power supply, a negative voltage drive power supply, a high-speed level conversion circuit, and a push-pull output stage. The positive voltage drive power supply provides a positive drive voltage, and the negative voltage drive power supply provides a negative turn-off voltage. The high-speed level conversion circuit is used to convert the low-voltage gate drive logic signal output in step 51 into a high-voltage control signal capable of driving the push-pull output stage. The push-pull output stage consists of a pair of complementary drive switches, whose output terminals are directly connected to the gate of the GaN switch transistor.

[0131] The forward drive voltage is preset based on the gate threshold voltage of the GaN switch and the gate overdrive voltage required for full conduction. For enhancement-mode GaN high electron mobility transistors, the typical gate threshold voltage is between 1.0V and 2.0V, while the typical forward drive voltage is between 5.0V and 6.0V, to ensure that the channel of the GaN switch is fully enhanced in the on-state and that the drain-source on-resistance reaches the minimum value specified in the datasheet. Simultaneously, the forward drive voltage is lower than the maximum rated gate voltage of the GaN switch, typically 7.0V, to prevent gate oxide breakdown damage.

[0132] Based on this, in step 533, according to the gate drive logic signal corresponding to the GaN switch transistor output in step 51 after soft-switching timing adjustment, at the moment when the gate drive logic signal flips from the off logic level to the on logic level, the high-speed level conversion circuit in the negative voltage direct drive gate drive circuit is controlled to convert the on logic level into a high-voltage control signal that drives the upper transistor of the push-pull output stage to turn on and the lower transistor to turn off, so that the positive voltage drive power supply provides a positive drive voltage pulse to the gate of the GaN switch transistor through the upper transistor of the push-pull output stage. Under the action of the positive drive voltage pulse, the gate voltage of the GaN switch transistor rises rapidly from the preset negative voltage level maintained in step 52 to the positive drive voltage. When the gate voltage exceeds the gate threshold voltage, the conductive channel of the GaN switch transistor begins to form and gradually strengthens, eventually entering a fully on state.

[0133] Since the turn-on time of the GaN switch is precisely aligned with the zero-voltage moment when its drain-source voltage naturally resonates to zero in step 514, under the zero-voltage turn-on condition, the drain-source voltage is approximately zero at the moment of turn-on. During the turn-on process, the superposition product of the drain-source voltage and the drain current, i.e. the turn-on loss power, is greatly reduced to close to zero, thereby achieving soft-switching turn-on with near-zero turn-on loss.

[0134] Based on this, in step 534, for the GaN switch whose drive type is shutdown drive as determined in step 531, according to the gate drive logic signal corresponding to the GaN switch output in step 51 after soft switching timing adjustment, at the moment when the gate drive logic signal flips from the turn-on logic level to the turn-off logic level, the high-speed level conversion circuit in the negative voltage direct drive gate drive circuit is controlled to convert the turn-off logic level into a high-voltage control signal that drives the upper tube of the push-pull output stage to turn off and the lower tube to turn on, so that the negative voltage drive power supply provides a negative turn-off voltage pulse to the gate of the GaN switch through the lower tube of the push-pull output stage.

[0135] Under the influence of the negative turn-off voltage pulse, the gate voltage of the GaN switch rapidly decreases from the positive drive voltage. After passing the gate threshold voltage, the conductive channel begins to clamp down. It continues to decrease until it reaches the negative drive power supply voltage, at which point the gate voltage stabilizes at a negative voltage level equal to the negative drive power supply voltage. Because the active Miller clamping circuit in step 52 synchronously clamps the gate voltage to a preset negative voltage level during the turn-off period of the GaN switch, the negative drive power supply voltage and the preset negative voltage level in step 522 can both be the same value. During the turn-off period, both maintain the negative gate voltage, forming a double negative voltage protection.

[0136] Since the turn-off time of the GaN switch is precisely aligned with the zero-current moment when its drain-source current naturally resonates to zero in step 514, under the zero-current turn-off condition, the drain current is approximately zero at the moment of turn-off. During the turn-off process, the superposition product of the drain-source voltage and the drain current, i.e. the turn-off loss power, is greatly reduced to close to zero, thereby achieving soft-switching turn-off with near-zero turn-off loss.

[0137] Based on this, in step 535, after the negative turn-off voltage pulse is applied in step 534, the active Miller clamping circuit in step 52 continues to maintain a low-impedance negative voltage clamp on the gate of the GaN switch. During the entire turn-off period, if the clamping capability of the active Miller clamping circuit becomes locally insufficient due to the excessive transient displacement current coupled by the Miller capacitor, the lower transistor of the push-pull output stage of the negative voltage direct drive gate drive circuit continues to conduct, providing a low-impedance path to the gate to provide negative voltage drive power, forming a parallel low-impedance clamping structure with the active Miller clamping circuit. Together, they stabilize the gate voltage at a preset negative voltage level, suppressing the gate voltage erroneous turn-on caused by the Miller effect.

[0138] This step applies gate drive pulses to all six GaN switches in the current switching cycle. GaN switches that need to be turned on receive a positive drive voltage pulse at zero voltage to achieve zero-voltage turn-on, while GaN switches that need to be turned off receive a negative turn-off voltage pulse at zero current to achieve zero-current turn-off. During the turn-off period, the gate voltage is maintained by the active Miller clamp circuit and the negative voltage direct drive gate drive circuit.

[0139] Step 54: According to the gate driving process, drive all GaN switches in the three-phase full-bridge inverter of the GaN device to alternately turn on and off within a preset switching frequency range of 50kHz to 200kHz, according to the final switching state combination; specifically including: Step 541: Based on the actual gate drive state of each GaN switch output in step 53, confirm that at the beginning of the current switching cycle, the gate voltages of all six GaN switches have reached their respective target levels, the gate voltages of the GaN switches that need to be turned on are stable at the positive drive voltage, and the gate voltages of the GaN switches that need to be turned off are stable at the preset negative voltage level.

[0140] Based on this, step 542 involves obtaining a preset switching frequency for the GaN device three-phase full-bridge inverter. The preset switching frequency is within the range of 50kHz to 200kHz. The lower limit of this range, 50kHz, is higher than the typical maximum switching frequency achievable by traditional silicon-based IGBT inverters due to switching losses and thermal management constraints, typically not exceeding 20kHz. This ensures the order-of-magnitude advantage of the GaN device three-phase full-bridge inverter in terms of switching frequency. The upper limit of this range, 200kHz, is limited by the engineering balance between the switching speed of the GaN switching transistors, the increase in switching losses, and the heat dissipation capacity, as well as the impact of parasitic inductance and capacitance within the SiP package module on the integrity of the gate drive signal. The specific value of the preset switching frequency is determined comprehensively based on factors such as the rated speed of the permanent magnet synchronous motor, the number of pole pairs, the expected bandwidth of the current loop, and the system's heat dissipation design.

[0141] Based on this, in step 543, the duration of the current switching cycle, i.e., the switching cycle time step, is determined according to the preset switching frequency obtained in step 542. The switching cycle time step is equal to the reciprocal of the preset switching frequency. For example, when the preset switching frequency is 100kHz, the switching cycle time step is 10 microseconds; when the preset switching frequency is 200kHz, the switching cycle time step is 5 microseconds.

[0142] Based on this, in step 544, according to the switching cycle time step determined in step 543 and the entire gate drive process described in steps 51 to 53, within the current switching cycle, all six GaN switching transistors in the three-phase full-bridge inverter are driven to alternately turn on and off according to the final switching state combination determined in step 44. At the end of the current switching cycle, the actual switching state of each GaN switching transistor is completely consistent with the final switching state combination determined in step 44.

[0143] In this step, because the GaN device three-phase full-bridge inverter operates at a high switching frequency of 50kHz to 200kHz, the duration of each switching cycle is only 5 to 20 microseconds. Compared to the 50 to 100 microsecond switching cycles corresponding to the traditional silicon-based IGBT solution at a switching frequency of approximately 10kHz to 20kHz, the duration of each switching cycle is shortened by several times. This high switching frequency characteristic significantly increases the voltage vector optimization update frequency of the MPC module in steps 41 to 43. The MPC module completes a complete ergonomic prediction, cost function calculation, and optimal vector selection within each switching cycle, and its update frequency is equal to the preset switching frequency. At a switching frequency of 200kHz, the MPC module completes a voltage vector update every 5 microseconds, and the discretization time step of current tracking is only 5 microseconds.

[0144] Meanwhile, the high switching frequency means that the frequencies of the switching frequency subharmonics and their sideband harmonics in the inverter output voltage are much higher than the fundamental operating frequency of the permanent magnet synchronous motor. For a typical servo motor with a rated speed of 3000 rpm and 4 pole pairs, its fundamental frequency is 200 Hz, while the lowest switching frequency of 50 kHz is already 250 times the fundamental frequency. Therefore, the inductive impedance of the motor stator winding has a very strong attenuation effect on the switching frequency subharmonic current component, effectively suppressing the switching ripple in the stator current, and making the output current waveform closer to an ideal sine wave.

[0145] After this step, all GaN switches in the GaN-equipped three-phase full-bridge inverter complete the alternating turn-on and turn-off of the current switching cycle according to the final switching state combination within the preset switching frequency range of 50kHz to 200kHz. The dual protection mechanism of active Miller clamping and negative voltage direct drive during the gate drive process ensures the safe and reliable operation of the GaN switches under high frequency and high voltage change rate conditions. Soft switching timing adjustment ensures that each switching action occurs under zero voltage or zero current conditions, minimizing the increase in switching losses caused by high frequency.

[0146] In a preferred embodiment of the present invention, step 6 includes: Step 61: Based on the alternating on and off actions of each GaN switch, determine the switching state transition sequence of each GaN switch within the current switching cycle; specifically including: Step 611: Based on the alternating on and off operation results of all six GaN switches in the three-phase full-bridge inverter output in Step 54 during the current switching cycle, obtain the actual switching operation type of each GaN switch during the current switching cycle. The actual switching operation type includes four cases: on-state switching from off to on-state, off-state switching from on-state to off-state, on-state holding action that remains on-state throughout the entire switching cycle, and off-state holding action that remains off-state throughout the entire switching cycle.

[0147] Based on this, in step 612, according to the final switching state combination determined in step 44 and the actual switching state of each GaN switch at the start of the current switching cycle output in step 54, the actual switching state of each GaN switch at the start of the current switching cycle is compared with the target switching state of that GaN switch in the final switching state combination. When the actual switching state of the GaN switch at the start of the current switching cycle is inconsistent with the target switching state, it is determined that the GaN switch has a switching state switching action within the current switching cycle. If the actual switching state is off and the target switching state is on, the switching state switching action is an on action; if the actual switching state is on and the target switching state is off, the switching state switching action is an off action. When the actual switching state of the GaN switch at the start of the current switching cycle is consistent with the target switching state, it is determined that the GaN switch has no switching state switching action within the current switching cycle.

[0148] Based on this, in step 613, according to the switching state switching actions of each GaN switch determined in step 612, the switching state switching sequence of each GaN switch in the current switching cycle is determined. The switching state switching sequence defines the switching action type of each GaN switch and its time coordinate in the current switching cycle, with the start time of the current switching cycle as the zero time reference point.

[0149] For a GaN switch that performs a turn-on action during the current switching cycle, its turn-on time is denoted as: ; For a GaN switch that performs a turn-off action during the current switching cycle, its turn-off time is denoted as: ; Among them, subscript Indicates the identifier index of the GaN switch. These correspond to the upper pipe of phase A bridge arm, the lower pipe of phase A bridge arm, the upper pipe of phase B bridge arm, the lower pipe of phase B bridge arm, the upper pipe of phase C bridge arm, and the lower pipe of phase C bridge arm, respectively. For the first The turn-on time of a GaN switch during the current switching cycle, in seconds (s). For the first The turn-off time of a GaN switch during the current switching cycle, in seconds (s). For step 513, the first The dead time delay of each GaN switch insertion is in seconds (s) and is a preset dead time interval. For step 514, the first When a GaN switch performs zero-voltage turn-on timing adjustment, the time offset relative to the end of the dead time delay is measured in seconds (s). This offset is determined by the resonant zero-crossing detection signal of the drain-source voltage of the GaN switch. For step 514, the first When a GaN switch performs zero-current turn-off timing adjustment, the time offset relative to the end of the dead time delay is measured in seconds (s). This offset is determined by the resonant zero-crossing detection signal of the drain-source current of the GaN switch.

[0150] For GaN switches that do not have a switching state transition during the current switching cycle, their current actual switching state remains unchanged throughout the entire switching cycle, with no defined turn-on or turn-off time.

[0151] Based on this, in step 614, according to the switching state transition sequence of each GaN switch determined in step 613, all switching action events occurring in the current switching cycle of each GaN switch are arranged in chronological order to generate a switching event sequence for the current switching cycle. The switching event sequence lists the occurrence time of each switching action event, the identifier of the GaN switch that performed the switching action, and the switching action type in ascending chronological order.

[0152] This step obtains the switching state switching timing of each GaN switch in the three-phase full-bridge inverter of GaN devices within the current switching cycle. This timing sequence clarifies the time coordinate of the turn-on or turn-off action of each GaN switch within the switching cycle, providing a timing reference for the soft switching timing coordination of the resonant network and the output of the stator winding drive voltage in subsequent steps.

[0153] Step 62: According to the switching state switching sequence, using the resonant network connected between the positive terminal of the DC bus of the three-phase full-bridge inverter and the midpoint of the three-phase bridge arm, before any GaN switch performs a turn-on operation, resonance is generated between the resonant inductor in the resonant network and the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch, causing the drain-source voltage of the GaN switch to naturally resonate to zero voltage, and the turn-on operation is performed at the zero voltage moment; specifically including: Step 621: Based on the switching event sequence output in step 61, identify the GaN switching transistors whose switching action type is turn-on in the current switching cycle, and determine these GaN switching transistors as the execution objects for zero-voltage turn-on.

[0154] Based on this, in step 622, for each GaN switch determined in step 621 that is subject to zero-voltage switching, a resonant network is obtained between the positive terminal of the DC bus of the three-phase full-bridge inverter for the GaN device and the midpoint of the bridge arm where the GaN switch is located. The resonant network includes a resonant inductor, a resonant capacitor connected in parallel with the drain and source of each GaN switch, and a DC bus capacitor connected between the positive and negative terminals of the DC bus. The resonant inductor is a passive inductive element with a fixed inductance value, which is pre-designed and determined according to the preset switching frequency and resonant period requirements. The resonant capacitor is a passive capacitive element with a fixed capacitance value, connected in parallel between the drain and source of each GaN switch, and its capacitance value is pre-designed and determined according to the output capacitance characteristics of the GaN switch and the resonant condition requirements of zero-voltage switching.

[0155] Based on this, in step 623, before the zero-voltage turn-on target GaN switch is about to perform the turn-on action, the change in the midpoint potential of the bridge arm caused by the turn-off action of the lower switch (for the turn-on of the upper switch) or the upper switch (for the turn-on of the lower switch) in the bridge arm where the GaN switch is located is used to excite the resonant inductor in the resonant network and the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch to generate LC series resonance.

[0156] In this step, the excitation process of LC series resonance is as follows: When the complementary switch of the bridge arm containing the GaN switch, i.e., another GaN switch in the same bridge arm as the GaN switch, switches from the on state to the off state, the output current path at the midpoint of the bridge arm is cut off. However, the inductive current in the motor stator winding cannot change abruptly. This inductive current is forced to redirect its flow through the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch, charging or discharging the resonant capacitor, causing its voltage, i.e., the drain-source voltage of the GaN switch, to begin to change. At the same time, a resonant current flows through the resonant inductor, and the magnetic field energy in the resonant inductor and the electric field energy in the resonant capacitor begin to exchange with each other, forming the LC series resonance process.

[0157] Based on this, in step 624, according to the LC series resonance process excited in step 623, the drain-source voltage of the GaN switch oscillates sinusoidally at the inherent resonant frequency of the LC series resonance under the resonance effect of the resonant capacitor and resonant inductor. The drain-source voltage starts to decrease (or increase) from the initial voltage value at the moment the complementary switch is turned off, and after a quarter of the resonance period, the drain-source voltage resonates to zero for the first time, forming the zero voltage moment.

[0158] In this step, the expressions for calculating the natural resonant angular frequency and natural resonant frequency of the LC series resonance are as follows: ; ; in, ω is the inherent resonant angular frequency of the LC series resonance, expressed in radians per second (rad / s). This is the inherent resonant frequency of the LC series resonance, expressed in Hertz (Hz). This is the inductance value of the resonant inductor, expressed in Henry (H). For parallel connection at the first The capacitance value of the resonant capacitor across the drain-source terminals of a GaN switch, measured in farads (F), includes the equivalent value of the GaN switch's own drain-source output parasitic capacitance and the externally applied parallel resonant capacitor.

[0159] The time elapsed from the moment the complementary switch turns off to the moment the drain-source voltage first resonates to zero, i.e., the resonant transition time, is equal to one-quarter of the resonant period. ,in It is the resonant period and The value of this resonant transition time is determined by the resonant inductor. and externally connected parallel resonant capacitor The design is pre-set so that the switching frequency is much smaller than the switching cycle time step throughout the entire preset switching frequency range (50kHz to 200kHz) to ensure that the timing conditions for zero-voltage turn-on can be reliably established within the current switching cycle.

[0160] Based on this, in step 625, according to the zero-voltage moment determined in step 624, combined with the turn-on moment of the GaN switch in step 613... Zero-voltage timing alignment offset determined by the resonant zero-crossing detection signal The drain-source voltage of the GaN switch is monitored in real time by a resonant zero-crossing detection circuit. The resonant zero-crossing detection circuit includes a voltage comparator. The non-inverting input of the voltage comparator is connected to a reference ground potential, and the inverting input is connected to the drain of the GaN switch through a voltage divider network. When the drain-source voltage drops from a positive value and crosses the zero voltage threshold, the comparator output flips, generating a resonant zero-crossing detection signal.

[0161] Based on this, in step 626, according to the resonant zero-crossing detection signal generated in step 625, the starting moment of the positive drive voltage pulse provided by the negative voltage direct drive gate drive circuit to the gate of the GaN switch in step 533, i.e., the moment when the gate drive logic signal flips from the off logic level to the on logic level, is precisely aligned with the zero voltage moment indicated by the resonant zero-crossing detection signal. At the zero voltage moment, the drain-source voltage of the GaN switch is approximately zero, and the positive drive voltage pulse provided in step 533 drives the GaN switch to conduct, achieving zero-voltage turn-on.

[0162] In this step, under zero-voltage turn-on conditions, the drain-source voltage of the GaN switch is approximately zero at the instant of turn-on. During the turn-on process, the drain current flowing through the GaN switch gradually increases from zero. The overlap area between the drain-source voltage and drain current in the time domain, i.e., the turn-on loss energy, is significantly reduced to near zero. The turn-on loss energy is calculated as the integral of the instantaneous product of the drain-source voltage and drain current over the turn-on transition time during each turn-on operation. Under zero-voltage conditions, since the drain-source voltage is zero at the start of turn-on, the integral result approaches zero, therefore the turn-on loss approaches zero.

[0163] This step completes the zero-voltage turn-on operation for all GaN switches that perform turn-on actions during the current switching cycle. Zero-voltage turn-on effectively reduces the turn-on losses of GaN switches under high-frequency switching conditions, ensuring that the switching losses of the inverter system do not increase linearly with the increase of the switching frequency when operating at high switching frequencies from 50kHz to 200kHz, thus solving the problem of increased losses caused by high-frequency operation.

[0164] Step 63: According to the switching state switching timing, before any GaN switch performs a turn-off operation, the drain-source current of the GaN switch is naturally made to zero through the resonant network, and the turn-off operation is performed at the zero current moment; specifically including: Step 631: Based on the switching event sequence output in step 61, identify the GaN switching transistors whose switching action type is turn-off in the current switching cycle, and determine these GaN switching transistors as the execution objects for zero-current turn-off.

[0165] Based on this, in step 632, for each GaN switch transistor determined in step 631 that is to perform a zero-current turn-off action, before the GaN switch transistor is about to perform a turn-off action, the drain-source current flowing through the GaN switch transistor is made to oscillate sinusoidally by utilizing the LC series resonance process between the resonant inductor in the resonant network and the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch transistor.

[0166] In this step, the excitation process of the sinusoidal oscillation of the drain-source current is as follows: At the end of the GaN switch's on-state, resonant energy exchange has already occurred between the resonant inductor in the resonant network and the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch. When the current direction in the resonant inductor is opposite to the drain-source current direction of the GaN switch, the resonant inductor current cancels out the load current carried by the GaN switch, causing the net current flowing through the channel of the GaN switch to gradually decrease.

[0167] Based on this, in step 633, according to the sinusoidal oscillation process of the drain-source current excited in step 632, the drain-source current flowing through the GaN switch oscillates sinusoidally at the inherent resonant frequency of the LC series resonance under the resonance effect of the resonant inductor and resonant capacitor. When the amplitude of the resonant inductor current increases to be equal to the amplitude of the load current carried by the GaN switch, the net current flowing through the channel of the GaN switch drops to zero, forming a zero current moment.

[0168] At zero current, no current flows through the conductive channel of the GaN switch. However, the GaN switch is still in the conducting state under the action of the gate positive drive voltage, and the channel remains in the enhanced state. The reason for zero current is that the external resonant network completely bypasses or cancels the load current, rather than the channel being pinched off.

[0169] Based on this, in step 634, according to the zero current moment determined in step 633, combined with the turn-off moment of the GaN switch in step 613... The zero-current timing alignment offset determined by the resonant zero-crossing detection signal. The drain-source current flowing through the GaN switch is monitored in real time by a resonant zero-crossing detection circuit. The resonant zero-crossing detection circuit includes a current sensing element and a zero-crossing comparator. The current sensing element is connected in series between the source of the GaN switch and the negative terminal of the DC bus, or the drain current is indirectly detected by a voltage drop detection scheme based on the on-resistance of the GaN switch. The zero-crossing comparator compares the output voltage of the current sensing element with a reference zero voltage. When the drain-source current drops from a positive value and crosses the zero current threshold, the comparator output flips, generating a current zero-crossing detection signal.

[0170] Based on this, in step 635, according to the current zero-crossing detection signal generated in step 634, the starting moment of the negative turn-off voltage pulse provided by the negative voltage direct-drive gate drive circuit in step 534 to the gate of the GaN switch is precisely aligned with the zero current moment indicated by the current zero-crossing detection signal. At the zero current moment, the negative turn-off voltage pulse provided in step 534 drives the GaN switch to turn off, achieving zero-current turn-off.

[0171] In this step, under zero-current turn-off conditions, the drain-source current of the GaN switch is approximately zero at the moment of turn-off. During the turn-off process, the drain-source voltage gradually rises from an approximately zero on-state voltage drop to the DC bus voltage. The overlap area between the drain-source voltage and drain current in the time domain, i.e., the turn-off loss energy, is significantly reduced to near zero. The formula for calculating the turn-off loss energy is the integral of the instantaneous product of the drain-source voltage and drain current over the turn-off transition time during each turn-off action. Under zero-current conditions, since the drain current is zero at the start of turn-off, the integral result approaches zero, therefore the turn-off loss approaches zero.

[0172] This step completes the zero-current turn-off operation for all GaN switches that performed turn-off actions during the current switching cycle. Zero-current turn-off effectively reduces the turn-off losses of GaN switches under high-frequency switching conditions, and together with the zero-voltage turn-on achieved in step 62, constitutes a complete soft-switching commutation process. In each switching cycle, the turn-on action of each GaN switch is performed under zero-voltage conditions, and the turn-off action of each GaN switch is performed under zero-current conditions, ensuring that the total switching losses, turn-on losses, and turn-off losses of the inverter system remain at a low level across the entire load range.

[0173] Step 64: Based on the turn-on operation at zero voltage and the turn-off operation at zero current, achieve zero-voltage turn-on and zero-current turn-off soft-switching commutation for each GaN switch, and output a high-frequency sinusoidal drive voltage to the stator three-phase windings of the permanent magnet synchronous motor through the midpoint of the three-phase bridge arm of the GaN device three-phase full-bridge inverter to drive the permanent magnet synchronous motor to rotate; specifically including: Step 641: Based on the zero-voltage turn-on and zero-current turn-off soft-switching commutation processes of each GaN switch completed in steps 62 and 63, it is confirmed that within the current switching cycle, all six GaN switches of the three-phase full-bridge inverter have completed the establishment and maintenance of the switching state according to the final switching state combination determined in step 44. The GaN switches that need to be turned on have completed zero-voltage turn-on and are in a fully conducting state, and the GaN switches that need to be turned off have completed zero-current turn-off and are in a fully turned-off state.

[0174] Based on this, in step 642, the potential of each midpoint of the three-phase bridge arm of the GaN device three-phase full-bridge inverter relative to the negative terminal of the DC bus is determined according to the final switching state combination of each GaN switch in the current switching cycle. Specifically, for phase A bridge arm, when the upper switch of phase A is on and the lower switch of phase A is off, the potential of the midpoint of phase A bridge arm is equal to the DC bus voltage; when the upper switch of phase A is off and the lower switch of phase A is on, the potential of the midpoint of phase A bridge arm is equal to the reference potential of the negative terminal of the DC bus. The same rule applies to determining the potentials of the midpoints of phase B and phase C bridge arms.

[0175] Based on this, in step 643, according to the three-phase bridge arm midpoint potential determined in step 642, the output voltages of phase A, phase B, and phase C of the GaN device three-phase full-bridge inverter to the stator windings of the permanent magnet synchronous motor are calculated. In this step, the phase voltage of each phase is defined as the potential difference between the midpoint of that phase bridge arm and the neutral point of the star-connected three-phase stator windings of the permanent magnet synchronous motor. When the three-phase stator windings are star-connected and the neutral point is not brought out, the phase voltage of each phase is obtained by calculating the instantaneous value of the three-phase bridge arm midpoint potential after deflection of the neutral point potential.

[0176] Based on this, in step 644, according to the three-phase output terminal voltage output in step 643, the corresponding phase voltages are applied to the stator A-phase winding, B-phase winding, and C-phase winding of the permanent magnet synchronous motor during the current switching cycle, generating three-phase drive currents in the three-phase stator windings. Since the GaN device three-phase full-bridge inverter operates at a preset switching frequency of 50kHz to 200kHz, the voltage pulse width and pulse timing output in each switching cycle are determined by the final switching state combination determined in step 44, which is directly mapped from the final voltage space vector selected in step 43.

[0177] In this step, due to the high switching frequency, the frequencies of the switching frequency subharmonics and their sideband harmonics in the inverter output voltage are much higher than the fundamental operating frequency of the permanent magnet synchronous motor. For a typical servo motor with a rated speed of 3000 rpm and 4 pole pairs, its fundamental frequency is 200 Hz, while the minimum preset switching frequency of 50 kHz is already 250 times the fundamental frequency. The inductive impedance of the motor stator winding has a very strong attenuation effect on the switching frequency subharmonic current component. The inductive reactance of the stator winding is proportional to the frequency, that is: ; in, The inductive reactance of the stator winding to harmonic frequencies is expressed in ohms (Ω). The frequency of the subharmonic of the switching frequency is expressed in Hertz (Hz). This is the equivalent inductance of each phase winding of the stator, measured in Henry (H). At 50kHz, the inductive reactance of the stator winding to this harmonic frequency is 250 times that to the 200Hz fundamental frequency. Therefore, the subharmonic current component of the switching frequency is suppressed to less than 0.4% of the fundamental current amplitude, the switching ripple in the stator current is effectively suppressed, and the output current waveform is closer to the ideal sine wave.

[0178] Based on this, in step 645, according to the three-phase drive current output in step 644, the three-phase drive current synthesizes a stator magnetomotive force vector rotating at an electric angular velocity in the stator winding of the permanent magnet synchronous motor. There is a spatial phase difference, i.e., an electromagnetic torque angle, between this stator magnetomotive force vector and the rotor magnetomotive force vector generated by the rotor permanent magnet. The two interact to generate electromagnetic torque, driving the rotor of the permanent magnet synchronous motor to rotate continuously at the actual rotor speed.

[0179] In this step, within each switching cycle, from acquiring the three-phase stator current and rotor position, to generating the q-axis current command in step 2, to executing current loop regulation and outputting the reference voltage vector in the stationary coordinate system, to the MPC module selecting the optimal voltage space vector and determining the final switching state combination, to driving each GaN switch to complete zero-voltage turn-on and zero-current turn-off through the gate drive circuit, and then to outputting the drive voltage to the stator winding and driving the motor to rotate in this step, the entire control link completes the closed loop within one switching cycle.

[0180] Because this invention adopts the FOC-MPC fusion architecture, the reference voltage vector output by the current loop regulator is directly fed into the MPC module for voltage vector optimization. The optimization result is directly mapped to the final switching state combination and drives the GaN switch to operate. There is no one- or two-step calculation delay and modulation delay introduced by the traditional SVPWM modulation stage between current deviation sampling and driving voltage application. The current loop voltage command can be applied to the motor stator winding without delay within the current switching cycle.

[0181] Based on this, in step 646, at the end of the current switching cycle, the control loop of the next switching cycle is entered. In the next switching cycle, based on the new three-phase stator current sampling values ​​and rotor position sampling values, all control processes from steps 1 to 6 are repeated to achieve continuous operation of the high-speed servo drive FOC-MPC fusion control based on GaN devices.

[0182] After this step, the permanent magnet synchronous motor rotates continuously under the high-frequency sinusoidal drive voltage output from the GaN device three-phase full-bridge inverter. The fundamental component frequency and amplitude of the motor stator current track the changes in the output commands of the position loop and speed loop in real time, achieving precise motion control. Due to the absence of modulation delay in the current loop, the online optimization of the MPC module cycle by cycle, the extremely small discretization error of current tracking under the high switching frequency of the GaN device, the soft-switching commutation ensuring operating efficiency under high-frequency conditions, and the active Miller clamp and negative voltage direct-drive gate drive circuit ensuring safe and reliable operation under high-frequency and high-voltage conditions, the entire servo drive system has achieved a comprehensive improvement over the traditional FOC plus modulation scheme in terms of position tracking accuracy, dynamic response speed, torque ripple suppression, system operating efficiency, and reliability.

[0183] A high-speed servo drive FOC-MPC fusion control system based on GaN devices includes: The acquisition module is used to acquire the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, it obtains the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system; based on the actual rotor speed and position command, it obtains the speed command; based on the speed deviation between the speed command and the actual rotor speed, it obtains the q-axis current command. The first calculation module is used to perform current loop adjustment on the d-axis current deviation and q-axis current deviation according to the q-axis current command, the actual d-axis current and the actual q-axis current, to obtain the d-axis reference voltage vector and the q-axis reference voltage vector; and to obtain the reference voltage vector in the stationary coordinate system by inverse coordinate transformation based on the d-axis reference voltage vector and the q-axis reference voltage vector. The second calculation module is used to select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; and to determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. The processing module is used to drive each GaN switch in the three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit according to the final switching state combination; and to output a drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm according to the switching state switching sequence of each GaN switch under the final switching state combination.

[0184] A computing device, comprising: Multiple processors; A storage device for storing a plurality of programs, which, when executed by a plurality of processors, cause the plurality of processors to implement the method.

[0185] A computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0186] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A high-speed servo drive FOC-MPC fusion control method based on GaN devices, characterized in that, Includes the following steps: Step 1: Obtain the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, obtain the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system. Step 2: Obtain the speed command based on the actual rotor speed and position command; obtain the q-axis current command based on the speed deviation between the speed command and the actual rotor speed. Step 3: Based on the q-axis current command, the actual d-axis current, and the actual q-axis current, perform current loop adjustment on the d-axis current deviation and the q-axis current deviation to obtain the d-axis reference voltage vector and the q-axis reference voltage vector; based on the d-axis reference voltage vector and the q-axis reference voltage vector, obtain the reference voltage vector in the stationary coordinate system through inverse coordinate transformation; Step 4: Select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. Step 5: Based on the final switching state combination, drive each GaN switch in the GaN device three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit. Step 6: Based on the switching sequence of each GaN switch under the final switching state combination, output drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm.

2. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 1, characterized in that, Step 1 includes: The instantaneous value of the three-phase stator current is acquired by a current sensor installed in the three-phase stator winding circuit of the permanent magnet synchronous motor at a sampling rate synchronized with the switching frequency of the three-phase full-bridge inverter of the GaN device, and the rotor mechanical angular position is acquired by an encoder installed at the rotor shaft end of the permanent magnet synchronous motor. Based on the instantaneous values ​​of the three-phase stator currents, the α-axis current components and β-axis current components in the α-β stationary coordinate system are obtained through Clark transformation. Based on the α-axis current component and the β-axis current component, and the electrical angle calculated based on the rotor mechanical angular position, the actual d-axis current and the actual q-axis current in the dq rotating coordinate system are obtained by Park transformation. The actual rotor speed is obtained by differentiating the rotor's mechanical angular position with respect to time.

3. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 2, characterized in that, Step 2 includes: Based on the actual rotor speed and the position command received from the upper-level controller, the position loop is proportionally adjusted to address the position deviation between the position command and the actual rotor position, thereby obtaining the speed command. Based on the speed deviation between the speed command and the actual rotor speed, a speed loop proportional-integral adjustment is performed to obtain the q-axis current command. At the same time, set the d-axis current command to zero.

4. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 3, characterized in that, Step 3 includes: The q-axis current deviation is calculated based on the difference between the q-axis current command and the actual q-axis current, and the d-axis current deviation is calculated based on the difference between the d-axis current command and the actual d-axis current, wherein the d-axis current command is zero. Based on the d-axis current deviation, perform d-axis current loop proportional-integral adjustment to obtain the d-axis reference voltage vector; Based on the q-axis current deviation, perform q-axis current loop proportional-integral adjustment to obtain the q-axis reference voltage vector; Based on the d-axis reference voltage vector and the q-axis reference voltage vector, and the electrical angle calculated from the rotor mechanical angular position, the α-axis reference voltage component and the β-axis reference voltage component in the α-β stationary coordinate system are obtained through inverse Park transformation, and used as the reference voltage vector in the stationary coordinate system.

5. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 4, characterized in that, Step 4 includes: Based on the reference voltage vector in the stationary coordinate system, within the current switching cycle of the GaN device three-phase full-bridge inverter, based on the discretized current prediction model of the permanent magnet synchronous motor in the stationary coordinate system, using the stator current sampling value at the current moment, the predicted current value corresponding to each basic voltage space vector of the GaN device three-phase full-bridge inverter is calculated for the next sampling moment. Based on the current tracking error between the predicted current value corresponding to each basic voltage space vector and the reference current value corresponding to the reference voltage vector in the stationary coordinate system, a cost function is constructed with the goal of minimizing the current tracking error. From all the basic voltage space vectors of the GaN device three-phase full-bridge inverter, select the final voltage space vector that minimizes the cost function; Based on the final voltage space vector, determine the final switching state combination of each GaN switch in the three-phase full-bridge inverter of the GaN device in the current switching cycle.

6. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 5, characterized in that, Step 5 includes: Based on the final switching state combination, generate gate drive logic signals corresponding to each GaN switch in the three-phase full-bridge inverter of the GaN device within the current switching cycle; According to the gate drive logic signal, for the GaN switch that is about to be turned on, the active Miller clamping circuit, which is integrated inside the SiP package module and arranged adjacent to the gate pad of the GaN switch, clamps the gate voltage to a preset negative voltage level during the gate turn-off period to suppress the gate voltage erroneously turned on due to the Miller effect caused by the drain-source voltage change rate. According to the gate drive logic signal, a negative voltage direct drive gate drive circuit cascaded with the active Miller clamp circuit provides a positive drive voltage pulse to the gate of the GaN switch during the gate turn-on period to turn it on, and provides a negative turn-off voltage pulse to the gate of the GaN switch during the gate turn-off period to turn it off. According to the gate driving process, all GaN switches in the three-phase full-bridge inverter of the GaN device are driven to alternately turn on and off within a preset switching frequency range of 50kHz to 200kHz, according to the final switching state combination.

7. The high-speed servo drive FOC-MPC fusion control method based on GaN devices according to claim 6, characterized in that, Step 6 includes: Based on the alternating turn-on and turn-off actions of each GaN switch, the switching state switching sequence of each GaN switch in the current switching cycle is determined. According to the switching state switching sequence, a resonant network connected between the positive terminal of the DC bus of the three-phase full-bridge inverter of the GaN device and the midpoint of the three-phase bridge arm is used. Before any GaN switch performs the turn-on operation, resonance is generated between the resonant inductor in the resonant network and the resonant capacitor connected in parallel across the drain-source terminals of the GaN switch, so that the drain-source voltage of the GaN switch naturally resonates to zero voltage, and the turn-on operation is performed at the zero voltage moment. According to the switching state switching timing, before any GaN switch performs a turn-off operation, the drain-source current of the GaN switch is naturally resonated to zero through the resonant network, and the turn-off operation is performed at the zero current moment. Based on the turn-on operation at the zero-voltage moment and the turn-off operation at the zero-current moment, the zero-voltage turn-on and zero-current turn-off soft switching commutation of each GaN switch is realized. Through the midpoint of the three-phase bridge arm of the three-phase full-bridge inverter of the GaN device, a high-frequency sinusoidal drive voltage is output to the stator three-phase winding of the permanent magnet synchronous motor to drive the permanent magnet synchronous motor to rotate.

8. A high-speed servo drive FOC-MPC fusion control system based on GaN devices, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to acquire the three-phase stator current and rotor position; based on the three-phase stator current and rotor position, it obtains the actual d-axis current, actual q-axis current, and actual rotor speed in the dq rotating coordinate system; based on the actual rotor speed and position command, it obtains the speed command; based on the speed deviation between the speed command and the actual rotor speed, it obtains the q-axis current command. The first calculation module is used to perform current loop adjustment on the d-axis current deviation and q-axis current deviation according to the q-axis current command, the actual d-axis current and the actual q-axis current, so as to obtain the d-axis reference voltage vector and the q-axis reference voltage vector. Based on the d-axis reference voltage vector and the q-axis reference voltage vector, the reference voltage vector in the stationary coordinate system is obtained through inverse coordinate transformation. The second calculation module is used to select the final voltage space vector from the basic voltage space vector of the GaN device three-phase full-bridge inverter based on the reference voltage vector in the stationary coordinate system; and to determine the final switching state combination of each GaN switch in the GaN device three-phase full-bridge inverter in the current switching cycle based on the final voltage space vector. The processing module is used to drive each GaN switch in the three-phase full-bridge inverter to alternately turn on and off within a preset switching frequency range through the gate drive circuit according to the final switching state combination; and to output a drive voltage to the stator winding of the permanent magnet synchronous motor through the midpoint of the bridge arm according to the switching state switching sequence of each GaN switch under the final switching state combination.

9. A computing device, characterized in that, include: Multiple processors; A storage device for storing a plurality of programs, which, when executed by the plurality of processors, cause the plurality of processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images